jrec(1) Paragon(TM) System jrec(1) NAME jrec - Processes the accounting log file entries and writes job records to stdout. SYNOPSIS jrec -s start_date -e end_date [ -j ] [ -v ] [ -h site_name ] [ -n total_nodes ] [ -f file ] [ -d file ] [ -p dir ] OPTIONS -j Produce a record for each job in a long listing format. By default, a summary report is produced. -v Produce status messages about jrec execution on the standard error output. -h site_name Specify the site_name used in the report output header. If omitted, the output will not include the system name or the column headings. -n total_nodes Specify the number of nodes on which the accounting report is based. Normally, this number represents the total available nodes in the .compute partition for the accounting period. If omitted, the default is the number of nodes on the current system, which is the largest bin specified in the actable file. -s start_date Specify the starting date of the report. -e end_date Specify the ending date of the report. Both start_date and end_date have the following format: mm/dd/yy [@hh[:mm[:ss]]] mm is the month, dd is the day, and yy is the last two digits of the year. The month and day can be specified with a single digit, when appropriate. For example, January 1 1995 can be specified as 1/1/95. -f file Specify a single input macdlog file for the report output. With the -f option, if the -s or -e options do not state the time, the reported time will be based on the data in the macdlog file. -d file Produce a downtime report in file in addition to the standard report sent to stdout. -p dir Specify an alternate log file directory. DESCRIPTION The jrec command processes the log files located in the /usr/spool/macs/log.d/macdlog.d directory and writes job records to stdout. The jrec command maintains an internal job table. As it reads the records from the log files, it creates a new job entry and records the job start time for every job initiation and writes a job record out for every job completion. If the job completion does not match any entry in the job table, the start time of the report or the last system uptime (whichever comes last) is used as the job start time. When a system-down record is encountered, a job record is written out for every job entry remaining in the job table. The system-down time is used as the job end time, and all job entries are deleted from the job table. In normal operation, the jrec command is invoked by the acctrep command. When invoked by itself, the jrec command outputs a text file that can be sorted and formatted to produce a report customized to match individual site requirements. If the command is invoked using the -h argument, the jrec output includes a user-specified site_name and the output also includes column headings that indicate column content. If the -j argument is used, jrec outputs the job records in a long format. Otherwise, jrec produces the report in a format that is easier for acctrep to process. EXIT STATUS 0 If the command completes successfully. 1 If the command fails. FILES /usr/spool/macs/conf/actable Accounting table information. /usr/spool/macs/log.d/macdlog.d/macd* The MACS log files. EXAMPLE # jrec -s 6/21/94 -e 6/21/94 -h SSD SSD Intel PARAGON Job Usage Report - produced on Tue Jun 21 09:28:22 1994 06/21/94 00:00:00 to 06/21/94 09:28:22 account login type size cpu-hours idle cpu-hrs under-used total charge ------- -------- ---- ---- ------------ ------------ ------------ ------------- swengr lynng 1 6 0.101 0.005 0.000 0.101 swengr stanh 1 6 0.101 0.010 0.000 0.101 quarks sydney 0 2 0.034 0.000 0.000 0.034 proj94 mmh 0 2 0.033 0.000 0.000 0.033 quarks sydney 1 6 0.034 0.207 0.000 0.034 quarks tonyc 1 6 0.102 0.005 0.000 0.102 (0772182000) 06/21/94 00:00:00 to (0772216102) 06/21/94 09:28:22 - down 0 seconds, scheduled down 0 seconds, scheduled 0, reboot 0; unscheduled reboot 0 In the jrec report, the parallel applications are sorted chronologically. The report summary contains the following information: account The MACS account's name. login The user who submitted the job. type The job type; either 0 for interactive jobs, 1 for NQS batch requests, or 2 for dedicated NQS batch requests. size The number of nodes (bin size) a parallel application ran on. If an application uses a different number of nodes than those allowed by the bins, the application is placed into the smallest bin that will hold it. The bin sizes are specified in the /usr/spool/macs/conf/actable file. cpu-hours The total node hours used by the job. idle cpu-hrs Idle time of an NQS batch request. Idle time is accrued if a batch request is rolled in but sits idle. No idle time is accrued when a NQS batch request is rolled-out. under-used Currently unreported (always 0). total charge The total charge for the job. jrec invoked with the -j option includes the following information in the report: account The MACS account's name. login The user who submitted the job. partition# A unique partition identifier that is used by the allocator when communicating with MACS. The same number appears in the MACS logfile entries for jobs running in that partition. size The number of nodes a parallel application ran on.typeThe job type; either 0 for interactive jobs, 1 for NQS batch requests, or 2 for dedicated NQS batch requests. cpu-hours The total node hours used by the job. rate First occurrence: the charge rate for the cpu-hours used. idle cpu-hrs Idle time of an NQS batch request. Idle time is accrued if a batch request is rolled in but sits idle. No idle time is accrued when a NQS batch request is rolled-out. rate Second occurrence: the rate charged for idle NQS requests. under-used Currently unreported (always 0). rate Third occurrence: the rate charged for under used node requests. total charge The total charge for the job. job submitted The date and time a job was submitted to NQS. These are listed as N/A for interactive jobs. job started The date and time an NQS job started. This is listed as N/A for interactive jobs. job ended The date and time an NQS job ended. This is listed as N/A for interactive jobs. st Job termination status. The value of this field is 1 if the job ran to completion, or 0 if it was terminated by a system crash or hang. In the event of a crash or hang, the application is assumed to have run until the crash or hang. reqst cpuhrs Either the time limit requested by the user with the qsub -lT option, or the per-request CPU limit for the queue. This is listed as N/A for interactive jobs. queue name The name of the queue an NQS request was submitted to. This will be listed as UNKNOWN for interactive jobs. SEE ALSO acctrep